Last updated: October 7, 2026, 11:58 AM IST
You open ChatGPT, you're ready to dig into a big research question, and... the Deep Research button is gone. Or it starts running and dies halfway. Or it just tells you you've hit a limit when you're pretty sure you haven't. If any of that sounds familiar, you're not alone — community threads about ChatGPT Deep Research not working have been piling up through 2026.
Here's the good news: in most cases, Deep Research isn't actually broken. It's limited, not missing. The button moved in a UI redesign, the quotas work differently than people expect, and a couple of quick checks fix the problem for the vast majority of users. This is a troubleshooting guide — start with the 60-second checks, then work down to the cause-by-cause fixes.
ChatGPT Deep Research Not Working? Start With These 60-Second Checks
Before you change anything, run through these. They solve most cases:
- Look under the "+" menu, then "More". Deep Research lives in the chat composer: click the + button, then look for "Deep research". After the August 2026 composer redesign, on many screens it sits inside a "More" submenu — this single UI change is the #1 cause of "the button disappeared" reports. It's not gone; it's nested.
- Check your in-product usage counter. OpenAI no longer publishes a fixed per-plan limit table — it now points everyone to the usage counter inside the product. If Deep Research errors instantly, your monthly quota is the prime suspect. Find your counter in Settings → Usage and see what's left.
- Try a different platform. Deep Research works on web, iOS, Android, macOS, and Windows. If the button is missing on web, check the mobile app (or vice versa) — rollouts and UI states differ between platforms.
- Hard-refresh and log back in. Boring, but stale sessions and cached UI states cause a real share of "missing feature" bugs. Ctrl+Shift+R (or the equivalent), then sign out and back in.
What Deep Research Actually Is (and What the Limits Are)
Quick refresher: Deep Research is OpenAI's autonomous research agent. You give it a complex question, it goes off and reads dozens of sources, and 5–30 minutes later you get a finished, cited report. There's also a lightweight version that runs on smaller, faster models with shorter answers — that's what free-tier users mostly get.
Now the limits — framed honestly, as of October 2026, because OpenAI has changed these more than once this year and no longer publishes a definitive table:
- Free / Go: ~5 lightweight queries a month. Enough to try it, not enough to rely on it.
- Plus ($20/mo): ~25 a month — roughly 10 full runs plus 15 lightweight.
- Pro ($200/mo): ~250 a month (about half lightweight). Pro ($100/mo) sits lower at roughly 5× Plus.
Treat every number above as a snapshot, not a contract — OpenAI's own docs now describe Pro tiers as "5×/20× Plus usage" instead of fixed task counts. Your in-product counter is the authority. Fello AI's current pricing breakdown tracks how these tiers compare if you want the wider context (ChatGPT free vs paid tiers on Fello AI).
Cause 1: Your Monthly Quota Is Exhausted
Symptoms: Deep Research errors immediately, or tells you to upgrade/wait. Everything else in ChatGPT works fine.
What's happening: You burned through your monthly allowance. This catches people off guard because the quota resets every 30 days from your first use — not on the 1st of the month. So "I haven't used it in weeks" can still mean "you're at 0 of 25."
The fix: Check your usage counter, note your reset date, and wait it out. To stretch next month's quota: use lightweight queries for scoping questions and save full runs for the final report. One more thing — a failed or abandoned run can still count against your quota, so don't fire off three parallel attempts when one is stuck.
Cause 2: The Button Moved (August 2026 Redesign)
Symptoms: "Deep Research used to be right there and now it's gone." You can still see it mentioned in help docs.
What's happening: The August 2026 composer redesign shuffled the menus. On many interfaces the option moved from a top-level button into + → More → Deep research. Mobile and desktop don't always match, either.
The fix: Click +, expand "More" (or the equivalent submenu on your platform), and look again. If you're on an older app version, update the app — some builds kept the old layout and some didn't, which is exactly why two people on the same plan see different menus.
Cause 3: Your Plan Tier Doesn't Include Full Deep Research
Symptoms: You can launch Deep Research, but reports are short, shallow, or labeled "lightweight" — or you're on Free/Go and only get a handful of runs.
What's happening: Free and Go tiers get a small number of lightweight queries. That's by design, not a bug. The lightweight version answers faster with shorter, less thorough reports.
The fix: If lightweight isn't cutting it for your work, the honest answer is that you need a higher tier — Plus gets the meaningful ~25/month allowance. But before upgrading, squeeze more out of what you have: write tighter, more specific prompts (a focused question produces a better lightweight report than a vague one), and do your broad scoping in regular chat before spending a Deep Research run on the final synthesis. For context on what each tier buys you these days, see our breakdown of the ChatGPT Pro $200 plan changes.
Cause 4: Regional or Rollout Differences
Symptoms: The feature works for a friend on the same plan but not for you. Or it worked last week and vanished.
What's happening: OpenAI rolls out UI changes and feature access gradually. Regional availability, A/B UI tests, and phased rollouts mean two accounts on identical plans can genuinely see different things.
The fix: Check the web app vs. the mobile app (they update on different schedules), make sure your app is fully updated, and give it a few days if a rollout is mid-flight. A VPN set to a different region is not a reliable fix — and it can cause account headaches — so don't go there.
Cause 5: Stuck or Failed Reports
Symptoms: A Deep Research run starts, spins for ages, then fails — or the report renders broken, or export doesn't work.
What's happening: This one is genuinely on OpenAI's side. Community threads from September 2026 document a wave of stuck runs and export failures — the long-running nature of the agent (5–30 minutes) makes it sensitive to server load and flaky connections (Deep Research reliability thread on the OpenAI community forum).
The fix: First, check whether it's actually stuck or just slow — these reports legitimately take up to 30 minutes. If it's dead, cancel and retry once (retries on server hiccups often succeed). Break huge questions into two smaller runs instead of one monster query. And export your finished reports immediately — don't trust the chat history as your only copy during a flaky week.
ChatGPT Deep Research Still Not Working: When to Contact Support
Run the quick checks, work through the causes, and if you're still stuck — a paid plan showing 0 of 25 used with the button completely absent everywhere — that's a genuine account-level issue. Contact OpenAI support from inside the app (Settings → Help) with your plan, platform, app version, and a screenshot. Include your usage counter reading; it saves a round of back-and-forth.
And a note on the "Free" experience: OpenAI has been moving more features behind quotas and ads this year — we covered the ChatGPT Finances free tier rollout too. If Deep Research limits feel tight on Free, that's the direction the product is heading, not a bug in your account.
Frequently Asked Questions
Is ChatGPT Deep Research gone in 2026?
No. It's "limited, not missing" for most users. The button moved into the + menu's "More" submenu after the August 2026 composer redesign, which is why many people think it disappeared.
How many Deep Research queries do I get?
As of October 2026: roughly 5 lightweight/month on Free/Go, ~25/month on Plus (about 10 full + 15 lightweight), and ~250/month on Pro. OpenAI no longer publishes a fixed table — your in-product usage counter is the final word.
When does the Deep Research quota reset?
Every 30 days from your first use — not on the 1st of the month. Check your usage counter for your specific reset date.
Why does Deep Research keep getting stuck?
Runs legitimately take 5–30 minutes, so give it time first. If it truly fails, cancel and retry once; server-side hiccups are documented in community threads from September 2026. Break very large questions into two smaller runs.
Can I use Deep Research on mobile?
Yes — it's available on web, iOS, Android, macOS, and Windows. If the button is missing on one platform, check another; UI rollouts differ between them.
Sources
- Fello AI — Is ChatGPT Free? (2026 tier comparison)
- OpenAI Community — Deep Research reliability discussion (Sep 2026)
My Personal Take
Most "Deep Research is broken" complaints I've seen come down to the moved button and the 30-day quota cycle. Once you know where to look — the + menu's More submenu — and when your own quota resets, the feature is usually fine. OpenAI could have communicated the redesign better; burying a flagship feature without telling anyone is how you get a hundred angry forum threads.
The real story, in my view, is the quotas. OpenAI quietly stopped publishing a fixed limit table, and the lightweight-versus-full split means "25 queries" doesn't mean what it used to. If Deep Research is part of your actual workflow, glance at your in-product counter weekly — and honestly, Plus at ~25 a month is the minimum serious tier. The lightweight version is a taster, not a tool.
0 Comments